Radar Emitter And Work State Identification

With the rapid development of the information revolution,especially in the field of electronic computers,electronic warfare has become the core of modern high-tech warfare.In recent years,the successful application of machine learning and artificial intelligence technology in various fields has led to more and more research in electronic warfare equipment and technology,especially in the field of radar countermeasures.Radar confrontation is mainly divided into radar electronic reconnaissance and radar electronic confrontation.Radar electronic reconnaissance is to detect and intercept the enemy's radar source signal to obtain information about its tactical and technical characteristics,and provide technical information support for battlefield situation assessment and enemy attack.Radar jamming is based on the reconnaissance and interferes with the enemy radar electronic equipment and system according to the battlefield situation to make it lose or reduce the performance.This paper mainly studies the unknown radar radiation emitter identification and radar working state in radar electronic reconnaissance,and the radar jamming decision problem in radar electronic countermeasures.Firstly,an unknown radar emitter identification method based on multi-source transfer learning is proposed for the identification of unknown radar emitters.This method combines transfer learning with multi-task learning.The effect and significance of the modulation mode and working mode of radar emitter parameters on the identification of radar radiation source models are analyzed.Then,the sample data to be tested is determined to be known data or unknown data by a detection method based on the local projection score.Finally,using the modulation characteristics and working mode of the emitter pulse data as an attribute,the unknown radar emitter identification model is learned by multi-source transfer based on transfer learning and multi-task learning by combining multiple known radar radiation source data,and the unknown radar source of radiation is identified.Then,a radar working state recognition method based on fusion feature is proposed for radar working state recognition.Firstly,the hierarchical structure of radar pulse data is introduced.A new definition and extraction method of "radar word" is proposed.The pulse amplitude data is introduced and analyzed.Then,based on the convolutional layer and the pooling layer of the CNN network in deep learning,the depth features of the pulse amplitude sequence data in the radar working state are extracted,and the time series depth features in the "radar word" sequence are extracted based on the RNN,and the two networks are obtained.The features are expanded and merged,and the working state is identified and classified by connecting the Softmax classifier.Finally,a radar joint jamming decision-making method based on Q-Learning algorithm is proposed for radar jamming decision-making.Firstly,the radar jamming problem analysis is modeled,the relationship between the radar working state and its threat level,the jamming pattern and jamming power in the radar jamming decision,and the radar jamming effectiveness evaluation problem are analyzed.Then the Q-Learning algorithm in reinforcement learning is applied to the radar jamming decision,the jamming pattern and the jamming power are taken as the joint jamming action,and the change of the radar threat level after the jamming is implemented as the basis for calculating the reward value,and the radar joint jamming decision model is established.Finally,by improving the learning rate in the Q value update formula to speed up the convergence of the model and reduce the oscillation to get the best jamming strategy.